Lateral connectivity as a scaffold for developing orientation preference maps
Introduction
Hubel and Wiesel first reported the phenomena of orientation selectivity in 1959 [14]. Since then, it has emerged as an important window through which we can observe cortical development. It has been shown in cats (see [9] for review) and ferrets [4] that normal development of orientation selectivity requires visual experience. With the advent of optical imaging [13], it has become possible to study the distribution of orientation preference across the surface of the visual cortex. Using this technique, it has been shown that the layout of orientation maps is stable and independent of visual experience. Orientation maps are present in binocularly deprived cats [7], [12] and are present in ferrets at eye-opening [5]. Chronic optical imaging studies have shown that maps are stable throughout the critical period [5], [12]. A dramatic reverse suture experiment [11] has shown that two eyes without common visual experience develop similar orientation maps.
The classic Hubel and Wiesel [15] model of orientation selectivity states that orientation selectivity of simple cells is a consequence of the geometrical arrangement of feedforward connections. A great deal of experimental evidence has accumulated in support of this model [6], [8], [20]. This implies that the stability of orientation maps is a consequence of the stability of the underlying feedforward connections. However, there is striking evidence that these feedforward connections are not stable during the critical period: they dramatically rearrange during ocular dominance segregation [19], and very brief periods of occlusion can lead to a massive rearrangement of the feedforward axons [1]. Therefore, these feedforward axons alone cannot account for orientation maps.
We propose that lateral connectivity establishes a “scaffold” for orientation maps. This provides a parsimonious explanation for the stability of orientation maps. It is established that horizontal connections in layer 2/3 form periodic “clusters” of finer axon branches that link columns of similar orientation selectivity [10], [21] (known as modular specificity). Bosking et al. [3] quantitatively assessed the specificity of horizontal connections with respect to both the orientation map and the map of visual space in the primary visual cortex in the tree shrew. They found that, in addition to displaying modular specificity, a neuron projects axons for longer distances and gives off more terminal boutons along the axis of visual space that corresponds to its preferred orientation (referred to as axial specificity). This finding, combined with evidence that horizontal connections are largely reciprocal [17], indicates that individual neurons receive input from other neurons whose receptive fields (RFs) are both co-oriented (of similar orientation preference) and co-axial (displaced along an axis in visual space that corresponds to their preferred orientation).
The key elements of our model are as follows:
- •
Anisotropic lateral connectivity, based on axial specificity, provides an experience-independent structural framework for the orientation map. This connectivity determines the layout of the map.
- •
Plastic feedforward connectivity, modeled by the BCM synaptic plasticity rule [2], provides the experience-dependent component of map development. This connectivity determines the sharpness of the orientation tuning.
Section snippets
Anisotropic lateral connectivity and naive networks
We use a two-eye, 32×32 neural network model. Fig. 1 shows the architecture for a 2×2 chunk of the network. Cortical neurons receive input about the visual environment via the feedforward connections and input from other cortical neurons via the lateral connections.
The incoming lateral connections to the network neurons are anisotropic. An example for one neuron is depicted in Fig. 2A. It can be shown that neurons with this kind of lateral connectivity and RFs that are shifted in visual space
Discussion
We have presented a model for the development of orientation preference maps in the visual cortex. The key elements of this model are anisotropic lateral connectivity, which provides a scaffold for the orientation map, and plastic feedforward connectivity, which gives rise to sharp orientation tuning. This model demonstrates how feedforward and lateral connectivity can work in concert to create stable maps that sharpen with visual experience, and provides a parsimonious explanation of the
Acknowledgments
This work was supported by NSF, ONR, and the Dana Foundation. D.H.G. was additionally supported by the UTRA program, the Royce Fellowship and the Goldwater Scholarship.
References (22)
- et al.
Strength and orientation tuning of the thalamic input to simple cells revealed by electrically evoked cortical suppression
Neuron
(1998) - et al.
Objective function formulation of the BCM theory of visual cortical plasticity: statistical connections, stability conditions
Neural Networks
(1992) - et al.
Cellular organization of reciprocal patchy networks in layer III of cat visual cortex (area 17)
Neuroscience
(1992) - et al.
Rapid remodeling of axonal arbors in the visual cortex
Science
(1993) - et al.
Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex
J. Neurosci.
(1982) - et al.
Orientation selectivity and the arrangement of horizontal connections in tree shrew striate cortex
J. Neurosci.
(1997) - et al.
Development of orientation selectivity in ferret visual cortex and effects of deprivation
J. Neurosci.
(1993) - et al.
Development of orientation preference maps in ferret primary visual cortex
J. Neurosci.
(1996) - et al.
The role of visual experience in the development of columns in cat visual cortex
Science
(1998) - et al.
Orientation selectivity of thalamic input to simple cells of cat visual cortex
Nature
(1996)
Development of neuronal selectivity in the primary visual cortex of the cat
Physiol. Rev.
Cited by (4)
Establishment of a scaffold for orientation maps in primary visual cortex of higher mammals
2008, Journal of NeuroscienceSimulation of the Visual Cortex with Laterally Connected Spiking Neural Networks
2003, Proceedings of the International Joint Conference on Neural NetworksStructured long-range connections can provide a scaffold for orientation maps
2000, Journal of Neuroscience